OVWST Meeting November, p. Evaluating Ocean Wind Products for Short-range Forecast Analyses of Wind-Driven Surface Currents John T. Gunn, Kathleen Dohan, Gary S. E. Lagerloef, and Fabrice Bonjean Earth & Space Research Seattle, Washington
Introduction OVWST Meeting November, p. Ocean Surface Currents Analyses Real processing system (OSCAR) is a satellite-derived surface current database based on a combination of quasi-steady geostrophic and locally wind-driven dynamics (Bonjean and Lagerloef, ). The geostrophic term is computed from the gradient of surface topography fields (AVISO/CLS). Wind-driven velocity components are computed from an Ekman/Stommel formulation with variable viscosity using QuikSCAT winds (FSU/COAPS) with a thermal wind adjustment using satellite SST data. Data available at http://www.oscar.noaa.gov.
Introduction OVWST Meeting November, p. 3 The aim of this study is to evaluate the accuracy of NCEP Reanalysis- and ECMWF ERA- Reanalysis GCM products at measuring ocean surface winds through comparisons with mooring wind data and QuikSCAT winds to begin to asses the feasibility of using GCM forecasts for OSCAR forecast currents. Results presented here: comparisons of the three winds with the TAO/TRITON and PIRATA mooring arrays comparisons of the currents with drifters using the different winds in OSCAR. The ultimate goal is to extend OSCAR to real- and short-range forecast currents on a -5 day base which includes faster wind-driven currents.
OSCAR LATEST DEVELOPMENTS OVWST Meeting November, p. Oscar Currents Sep m/s Oscar Currents Sep 5 o N 5 o N m/s.. o N.5 o N.5 39 o N. 39 o N..3.3 3 o N. 3 o N. 33 o N m/s 7 o W 7 o W o W o W 5 o W o W. 33 o N m/s 7 o W 7 o W o W o W 5 o W o W. Newest development of OSCAR: increase resolution from one degree to /3 degree. Smaller-scale features in higher resolution model, e.g. Gulf Stream region. Much closer to coasts with optimal coverage at each step. Improved model for equatorial currents.
COMPARISON WITH DRIFTERS OVWST Meeting November, p. 5 Currents are interpolated onto the drifter locations (averaged over day). Zonal and meridional currents vs drifter velocities. We want to incorporate faster wind-driven mixed layer currents than our current quasi-steady model allows. This requires an extension from 5-day to -day base and -dependent terms in the momentum equations.
QuikSCAT Comparisons with moorings OVWST Meeting November, p. Time series of mooring at (5 o W, o N) and QuikSCAT data over months. U (top) and V (bottom) Mooring (blue) and QuikSCAT (black). U (m/s) V (m/s) QuikSCAT Zonal Winds vs TAO for year at mooring 5w n QuikSCAT Mooring / /9 / /3 /3 / /3 / /7 QuikSCAT Meridional Winds vs TAO for year at mooring 5w n / /9 / /3 /3 / /3 / /7
All winds comparisons with moorings U (m/s) V (m/s) QuikSCAT Zonal Winds vs TAO for year at mooring 5w n QuikSCAT Mooring / /9 / /3 /3 / /3 / /7 QuikSCAT QuikSCAT Meridional Winds vs TAO for year at mooring 5w n / /9 / /3 /3 / /3 / /7 U (m/s) NCEP Zonal Winds vs TAO for year at mooring 5w n NCEP / /9 / /3 /3 / /3 / /7 U (m/s) V (m/s) ECMWF Zonal Winds vs TAO for year at mooring 5w n ECMWF Mooring / /9 / /3 /3 / /3 / /7 ECMWF ECMWF Meridional Winds vs TAO for year at mooring 5w n / /9 / /3 /3 / /3 / /7 NCEP Mooring V (m/s) NCEP Meridional Winds vs TAO for year at mooring 5w n / /9 / /3 /3 / /3 / /7 OVWST Meeting November, p. 7
All winds comparisons with moorings U (m/s) QuikSCAT Zonal Winds vs TAO for year at mooring 5w n QuikSCAT Jan Feb Mar Apr May Jun Jul QuikSCAT Mooring U (m/s) ECMWF Zonal Winds vs TAO for year at mooring 5w n ECMWF Jan Feb Mar Apr May Jun Jul ECMWF Mooring V (m/s) QuikSCAT Meridional Winds vs TAO for year at mooring 5w n Jan Feb Mar Apr May Jun Jul U (m/s) NCEP Zonal Winds vs TAO for year at mooring 5w n NCEP Jan Feb Mar Apr May Jun Jul V (m/s) ECMWF Meridional Winds vs TAO for year at mooring 5w n Jan Feb Mar Apr May Jun Jul NCEP Mooring V (m/s) NCEP Meridional Winds vs TAO for year at mooring 5w n Jan Feb Mar Apr May Jun Jul OVWST Meeting November, p.
OVWST Meeting November, p. 9 Correlation QuikSCAT U CORRELATION day 5 5 3 35.5 QuikSCAT V CORRELATION day 5 5 3 35.5 ECMWF U CORRELATION day 5 5 3 35.5 ECMWF V CORRELATION day 5 5 3 35.5 NCEP U CORRELATION day 5 5 3 35 longitude ( o E).5 NCEP V CORRELATION day 5 5 3 35 longitude ( o E).5 Correlations with each mooring. The products are interpolated onto each mooring location at the mooring s. Statistics are calculated from -Jul-999 to 3-Aug-. U component on the left, V on the right. QuikSCAT, ECMWF, NCEP
OVWST Meeting November, p. SKILL QuikSCAT U SKILL day 5 5 3 35 QuikSCAT V SKILL day 5 5 3 35 ECMWF U SKILL day 5 5 3 35 ECMWF V SKILL day 5 5 3 35 NCEP U SKILL day 5 5 3 35 longitude ( o E) NCEP V SKILL day 5 5 3 35 longitude ( o E) SKILL = std(u mooring U product ) std(u mooring )
OVWST Meeting November, p. Relative Standard Deviation QuikSCAT U RSTD day 5 5 3 35 QuikSCAT V RSTD day 5 5 3 35 ECMWF U RSTD day 5 5 3 35 ECMWF V RSTD day 5 5 3 35 NCEP U RSTD day 5 5 3 35 longitude ( o E) NCEP V RSTD day 5 5 3 35 longitude ( o E) Relative standard deviation is a measure of the comparison between variances of products. RSTD = std(u mooring) std(u product ) std(u mooring )
OVWST Meeting November, p. Overall Score T = day CorrU CorrV SkillU Skill V Rstdu Rstdv ScoreU ScoreV QuikSCAT.9.9.57.5 -. -. 7.. ECMWF.9.7.5.9.. 7.3.3 NCEP.5.55 -.7 -. -.5 -. 3. 3.3 T = 5days CorrU CorrV SkillU Skill V Rstdu Rstdv ScoreU ScoreV QuikSCAT.9.93.5.57 -. -.5 7.3.59 ECMWF.95.9.9..7.3 7. 7.39 NCEP.7.77..7 -.7 -.3.9.97 Final score (out of ) reflects the overall ability to match the mooring data. SCORE = ( + COR) ( + SKILL) + RSTD.5 T = 5 days indicates statistics for data mapped onto OSCAR base: 5-day and smoothed over days.
TOWARDS FASTER TIMESCALES OVWST Meeting November, p. 3 OSCAR currents are calculated on a -day base using -day winds with interpolated SSH and SST values. No acceleration term included in the model. Sample output. Comparison with drifters: Data from drifters in the box are averaged over one day. Model currents are interpolated onto the drifter locations at each day. Statistics are calculated on the ensemble of points during -Mar- to 3-May-.
COMPARISON WITH DRIFTERS N=37 rmsd=5cm/s rdstd=7% md=.3cm/s cor=.3 sk=. sco=3.3 N=37 rmsd=cm/s rdstd=% md=.3cm/s cor=. sk=. sco=3.5 Zonal and meridional OSCAR currents using QuikSCAT winds vs drifter velocities. Results are very similar between winds. Corr Skill Score Zonal OSCARQSCD vel (cm/s) 5 Meridional OSCARQSCD vel (cm/s) u.3. 3.35 v.. 3.53 5 5 5 Zonal Drifter vel (cm/s) 5 5 Meridional Drifter vel (cm/s)........ Frequency (no unit)... Frequency (no unit)......... 5 5 5 5 Zonal [OSCARQSCD Drifter] vel (cm/s) (bar width = 5 cm/s) 5 5 5 5 Meridional [OSCARQSCD Drifter] vel (cm/s) (bar width = 5 cm/s) OVWST Meeting November, p.
COMPARISON WITH DRIFTERS OVWST Meeting November, p. 5 T = day CorrU CorrV SkillU Skill V Rstdu Rstdv ScoreU ScoreV QuikSCAT.3....7. 3.35 3.53 ECMWF..7...7. 3.37 3.5 NCEP..3.9..7. 3. 3.5 T = 5days CorrU CorrV SkillU Skill V Rstdu Rstdv ScoreU ScoreV QuikSCAT.57.3..3.9. 3. 3.3 Better results for increased resolution is promising for extending surface currents beyond the quasi-steady 5-day model.
Future Work OVWST Meeting November, p. We have presented a preliminary assessment of QuikSCAT, ECMWF and NCEP winds for the purpose of an improved wind-driven component to the OSCAR system. More analyses will be performed, in particular spatial comparisons (e.g. comparisons of EOF structures) and evaluation of forecast products. Goals: include -dependent dynamics in OSCAR to include high-frequency wind-driven currents extend OSCAR capability to nowcast and forecast improve the wind-driven turbulent mixing scheme.